The tool that reduced nurse overtime shifts by 85% during a 6-week window
September 2022
Problem & Motivation
Creating nurse staffing schedules is a constant battle between meeting uncertain patient demand and improving nurse satisfaction. Too often, poor planning or unprecedented events will result in overtime shifts, insufficient staffing, unmet patient needs, and employee exhaustion. H2O’s Nurse Staffing Schedules solution addresses each one of these challenges by providing multiple customized and optimized scheduling plans that hospital managers can compare, evaluate, adjust, and deploy.
H2O’s algorithm was tested in partnership with one of the largest hospitals in the US to optimize nurse staffing schedules taking into account 3 different nurse shift types, 11 different positions, and individual preferences, as shown in the case study below. The results were an overall reduction of 5% in undesirable shift patterns, including an 85% reduction in overtime shifts and a 17% reduction in weekend shifts per day. H2O’s solution to nurse scheduling also promoted greater job satisfaction due to increased diversity of staffing cycles and greater support from new trainees.
Two-Step Optimization Approach
The Nurse Staffing Schedules solution is broken down into two important steps. First, the foundational machine learning algorithm will analyze and interpret historical demand data provided by each hospital. Then, it will generate a set of optimized aggregate and granular scheduling solutions that prioritize individual nurse preferences, while also meeting patient demand.
Step 1 Model: Aggregate Staffing Levels
In order to effectively analyze and interpret historical patient demand data, the machine learning algorithm is given a clear objective, in this case, to minimize the total number of shifts scheduled without sacrificing the number of patients served. Having a straightforward objective means that the constraints, controls, and overall factors going into the algorithm will also be highly relevant and more accurate. The H2O Nurse Staffing Schedules solution primarily utilizes data from the most recent 6 weeks of demand, while putting more emphasis on the current trends. The algorithm outputs a projected 6-week calendar, specifying the number of nurses needed each day, by tier, by department, by shift, i.e. night, evening, day. The schedule takes into account nurse individual preferences for the next 6-8 weeks.
Step 2 Model: Individual Nurse Scheduling
After analyzing historical demand data, the algorithm optimizes nurse staffing schedules by heavily weighing individual preferences to reduce weekend, holiday, and overtime shifts and maximize nurse satisfaction. In a successful product test, the H2O solution provided daily assessment of 100+ individual nurses that work and train on 3 different shifts at 11 separate positions. A task like this is extremely complicated and time consuming for hospital managers to do manually, and especially at scale. The automated H2O solution makes this process quick, painless, and effective.
User Interface
1. Solve optimization model with choice of parameters
2. Compare and select alternative solutions
3. View, edit, and export final schedule
Impact
Vigorous testing of the H2O Nurse Staffing Schedule solution resulted in the reduction of overtime shifts by 85%, overall reduction of all shifts by 5%, and a wide range of career benefits to individual nurses, outlined below.
